First attempt at using TL494 inverter circuit, periodic "shoot-thru" on power-up...

Thread Starter

Hamlet

Joined Jun 10, 2015
519
(Sorry for the poor schematic, but this is what I'm working with.)

This is just my first go-around with this chip, the TL494. I realize this isn't very
sophisticated, no output feedback, soft start, etc., I'm just trying to get my feet wet.

My problem is when I power-up, sometimes it's behaved, with low quiescent current (0.5A),
and other times it seems to shoot-thru on power up, and maxes out my power supply at 11A.

It doesn't matter if I have a load on it or not. It seems to happen 50% of the time, but not neccesarily
every other instance. I have three decoupling/supply caps on the Vcc pins. Mosfets are STP90NF03L,
N-channel 30V - 0.0056Ω -90A, rated for switching regulators.

When it works, with a 10w bulb for a load, it draws about 2A. Output voltage is 145v with my 120/8-0-8v transformer.

One last item, it isn't on the schema, but I added 10k pull down resistors on the gates. I don't think it needs them,
but it didn't help me none either...

1612484125537.png
 

Alec_t

Joined Sep 17, 2013
14,280
Pins 9 and 10 are providing active pull-ups on the MOSFET gates, but without active pull-downs the MOSFETS will switch off slowly via the pull-down resistors, so there will be a lot of 'shoot-through' (i.e. both MOSFETS on at the same time). I don't see why that would be happening only 50% of the time; I would expect it every cycle.
 

DickCappels

Joined Aug 21, 2008
10,152
Please tell me that you are NOT really using 10 uf as Ct.

This chip was designed to drive bases of bipolar power transistors, not the gates of MOSFETs. Without an active pull-down you are at the mercy of a pull-down resitor to discharge the massive 2700 pf input capacitance.

If you really are using a 10 uf capacitor as Ct, the the frequency will be too low to use a "normal" switching supply transformer you might be seeing the core saturating.
 

Ian0

Joined Aug 7, 2020
9,671
If you really need to use a vintage component, then the SG3525 is equally ancient but far better equipped for driving MOSFETs.
 

Thread Starter

Hamlet

Joined Jun 10, 2015
519
Please tell me that you are NOT really using 10 uf as Ct.

This chip was designed to drive bases of bipolar power transistors, not the gates of MOSFETs. Without an active pull-down you are at the mercy of a pull-down resitor to discharge the massive 2700 pf input capacitance.

If you really are using a 10 uf capacitor as Ct, the the frequency will be too low to use a "normal" switching supply transformer you might be seeing the core saturating.
Thanks for the reply, Dick. I like to learn by doing, but that's not to say I didn't read thru the datasheet a dozen times... I missed the active pull-down bit until after dinner, but after I ordered and received the part. Yes, I'm using 10uF for CT, and I've tried 1uF, and .1uF. I replaced Rt with a 10K pot. I have a selection of iron-core power transformers on the shelf, so yeah, I wanted a low freq.

I lowered my gate resistors to 68ohms, and the situation improved, but not completely. I think I need some kind of
active-pull-down, a mosfet driver... maybe something like this will work:

1612543678010.png
???
 

crutschow

Joined Mar 14, 2008
34,285
yeah, I wanted a low freq.
How low?
Are you trying to generate a 50/60Hz output?

Note that a transformer designed for a specific sine-wave voltage and frequency can saturate with the same frequency and peak-voltage square-wave input, as you have.
To prevent that the square-wave peak (equal to the RMS) transformer voltage should be no more that the sine-wave RMS voltage
Thus for a 12V supply, the winding should be rated for 12Vrms, not 8Vrms as you have.
 
Last edited:

Thread Starter

Hamlet

Joined Jun 10, 2015
519
How low?
Are you trying to generate a 50/60Hz output?
Note that a transformer designed for a specific sine-wave voltage and frequency can saturate with the same frequency and peak-voltage square-wave input, as you have.
Crutschow! Okay, I seem to remember that from somewhere... and that current draw explodes when
saturation is reached. And no, I am not specifically aiming for 60Hz, but something under 1kHz, because
it's an iron core.

Just now, I went back to my first schema and added 100ohm pull-down resistors on the gates. My resistor
values might be a little low, but the TL494 is not heating, even though I am at the edge of the spec.

I am getting reliable operation now, can drive halogen lamps, or transformer without shoot-thru. I have
slighty elevated quiescent current (0.6A), but with the freq. dialed up to 400Hz, my output loads are operating about ten percent more efficient. But at anything over 900Hz and the shoot-thru comes back (might even be saturation).
 

crutschow

Joined Mar 14, 2008
34,285
But at anything over 900Hz and the shoot-thru comes back (might even be saturation).
If you are referring to transformer saturation, that goes down as the frequency increases.
It's determined by the transformer magnetizing current which is proportional to voltage and inversely proportional to frequency.
 

Thread Starter

Hamlet

Joined Jun 10, 2015
519
If you are referring to transformer saturation, that goes down as the frequency increases.
It's determined by the transformer magnetizing current which is proportional to voltage and inversely proportional to frequency.
Ah, that's why I get improvement with increased frequency, thanks.

I just did a test, with an input current of 11A@12v=132w, I get an output of 130v @ .83A, or 108w.
That's about 82% efficiency. I'm happy with that for my first time out.
 

Ian0

Joined Aug 7, 2020
9,671
For a 50Hz transformer, you need at least 70Hz if you are driving it with a square-wave of the same peak voltage. The efficiency will go down with increased frequency because of increased core losses, and because you then have far too many turns and too much inductance.
The same sized core can be used at 400Hz, but it will need thinner laminations and fewer turns, and then you can get an impressive amount of power out of it. Also bear in mind that any magnetostriction will make it hum at a frequency at which your ears are much more sensitive!
There's a reason that 400Hz is used as the AC power frequency in aircraft. I assume that above 400Hz the core losses start to outweigh any increase in efficiency you may gain
 

Thread Starter

Hamlet

Joined Jun 10, 2015
519
For a 50Hz transformer, you need at least 70Hz if you are driving it with a square-wave of the same peak voltage. The efficiency will go down with increased frequency because of increased core losses, and because you then have far too many turns and too much inductance.
The same sized core can be used at 400Hz, but it will need thinner laminations and fewer turns, and then you can get an impressive amount of power out of it. Also bear in mind that any magnetostriction will make it hum at a frequency at which your ears are much more sensitive!
There's a reason that 400Hz is used as the AC power frequency in aircraft. I assume that above 400Hz the core losses start to outweigh any increase in efficiency you may gain
Almost like magic. I'm not adverse to rewinding it, but I can't do anything about lamination thickness. I have some large ferrite transformer cores, I may switch to them and see what I can learn.

I may need to parallel mosfets for more power; does my schematic in post #5 look workable?
 

DickCappels

Joined Aug 21, 2008
10,152
The circuit in Post #5 is a good starting place. You can reduce R5 and R6 depending upon the types of transistors used in the gate drive buffers you can probably make them anywhere between 47 ohms and 100 ohms. 500 ohms may slow things down and increase dissipation.
 

Thread Starter

Hamlet

Joined Jun 10, 2015
519
Yes. Will you be using the error amps in the chip for something, e.g. current limiting?
I may. I've looked and datasheets from Onsemi, Ti, etc., and they are not the easiest to decipher.
Ti shows an ap-note, but my downloads time-out, so no luck there. Eventually, I will employ
PWM on chip to fine-tune the target voltage, and after that develop some kind of feedback thru an
opto. If I can create a workable scheme, I anticipate using this with an output bridge to
run fixed-load d.c. motors at 90v/180v* (eg. agricultural spray pumps, hydraulics, irrigation.)

*Fractional HP d.c. motors at voltages of 90/180v are cheap, as opposed to 12/24v, and the cabling/control is
more manageable.
 

Thread Starter

Hamlet

Joined Jun 10, 2015
519
The circuit in Post #5 is a good starting place. You can reduce R5 and R6 depending upon the types of transistors used in the gate drive buffers you can probably make them anywhere between 47 ohms and 100 ohms. 500 ohms may slow things down and increase dissipation.
Okay, made changes. I'll probably go with BD139/BD140 for the totem drivers. I'll increase
the frequency considerably if I go with ferrite transformers.
(SG3525 is looking to be the better route, I have them on order.)

1612640640428.png
Edit 2x
 

Attachments

Last edited:

Ian0

Joined Aug 7, 2020
9,671
Hang on a minute - which way round does the dead time work on this IC? I think that its built-in transistors are intended as power output transistors in low power applications, so I think that the transistors are switched so that during the dead-time they are both off.
So if you use the internal transistors as pull-downs to switch the MOSFETs off, and 500 ohm resistors as pull-ups to switch the MOSFETs on, the you have overlap where both power transistors are on, and we have ensured the shoot-through you were trying to avoid.
I still say you'd be better off with the SG3525, in which case you wouldn't need the four extra drive transistors.
 

Thread Starter

Hamlet

Joined Jun 10, 2015
519
Hang on a minute - which way round does the dead time work on this IC? I think that its built-in transistors are intended as power output transistors in low power applications, so I think that the transistors are switched so that during the dead-time they are both off.
So if you use the internal transistors as pull-downs to switch the MOSFETs off, and 500 ohm resistors as pull-ups to switch the MOSFETs on, the you have overlap where both power transistors are on, and we have ensured the shoot-through you were trying to avoid.
I still say you'd be better off with the SG3525, in which case you wouldn't need the four extra drive transistors.
Good catch Ian! Schema in post #5 now updated with 120ohm pull down resistors (R8 &r9).
 
Top